Jeronimo Castrillon
Programming models and abstractions for computational efficiency
The demise of scaling laws in micro-electronics has led to an era of innovation in software and hardware architectures aimed at improving the energy efficiency of computing systems. This talk discusses programming models and software abstractions that help cope with demanding applications with stringent constraints on highly heterogeneous and emerging computing systems. Concretely, we describe dataflow and actor-based abstractions for energy-efficiency computing and a reactive extension thereof for time-deterministic execution of cyber-physical systems. We then turn to high-level programming abstractions and compilation flows for emerging near and in-memory computing systems. We describe current efforts on building an extensible framework around the MLIR compiler infrastructure to abstract from individual technologies to foster re-use, targeting commercial near-memory systems, in-memory cross-bars, computing with content-addressable memories and bulk-wise logic operations.
back to overview
